1. Field of the Invention
The present invention pertains generally to coded data generation or conversion and more particularly to systems and methods for conversion between analog and digital signals.
2. Description of Related Art
Transducers (e.g., speakers and microphones) have long been designed and manufactured for systems such as telephones, televisions, recorders, and stereo equipment. However, the recent proliferation of digital systems, particularly computers, that incorporate audio and/or telephony capabilities has increased demand for low cost digital-to-analog (D/A) and analog-to-digital (A/D) converters that provide high sound quality. For example, the integration of computers and audio features, as implemented in personal computer sound cards, has found popularity in education, gaming, teleconferencing, voice recognition, and many other applications. Because these applications often target home, school, and small office markets, low cost converter implementations are crucial to their success in the market. Consequently, improvements in sound quality that limit, or preferably do not require, additional components are important in these markets.
Analog and digital signal conversion is typically implemented using a codec (coder/decoder), which is a device that converts analog signals to digital to be read or stored by a computer and that converts the digital signals back to analog. Sound cards and video cards typically use this kind of codec. For example, a personal computer sound system typically requires conversion of analog signals received from a microphone into digital data for processing and storage. Likewise, such a system typically requires conversion of digital data generated by the system into analog signals for driving an audio speaker. In another example, telephone systems often incorporate both analog and digital signal processing and therefore are required to convert between analog and digital signals.
FIG. 1 depicts a typical transducer system. Microphone 104 exhibits an exemplary frequency response 101, and speaker 102 exhibits an exemplary frequency response 100. Frequency response is a term that characterizes the ratio of the amplitude of an output signal to the amplitude of an input signal (i.e., gain) over a range of frequencies. In general, sound waves are detected by the microphone 104 and an analog signal representing the input sound waves is communicated to A/D and D/A codec 106 via signal path 112. Codec 106 converts the analog signal to a digital signal, which is received by digital audio system 108 via signal path 114. In the opposite direction, a digital signal representing output sound is generated by digital audio system 108 and communicated to codec 106 via signal path 114. It should be understood that signal path 114 may share a single physical connection or be distributed among more than one connection. Codec 106 converts the digital signal to an analog signal and communicates an analog signal to speaker 102 via signal path 110.
A popular technique used in A/D and D/A conversion applications is called "sigma-delta" modulation, which is also often referred to as "delta-sigma" modulation. The terms are virtually interchangeable, with a distinction possibly being made only when the relative position of the summing (sigma) block within the modulator is being emphasized. Sigma-delta modulation is a preferred method of conversion between digital and analog signals because it achieves high-resolution by precise timing instead of by precisely-matched on-chip components, such as the resistors or capacitors used in resistive or capacitive divider techniques.
Sigma-delta codecs, which may include one or more sigma-delta modulators and filter stages, are also called oversampling converters or noise shaping converters. Oversampling and noise shaping evoke the two basic principles involved in the A/D operation of sigma-delta modulators. Oversampling spreads the quantization noise power over a bandwidth equal to the sampling frequency, which is much greater than the signal bandwidth. In noise shaping, the modulator behaves as a lowpass filter on the signal, and as a highpass filter on the noise, thus "shaping" the quantization noise so that most of the noise energy will be above the signal bandwidth. That is, the quantization noise is primarily distributed at frequencies above the relevant frequency band of the signal. A digital lowpass filtering stage, typically called a decimation filter, then removes the out-of-band quantization noise and downsamples the digital signal to the Nyquist rate.
In D/A conversion, a sigma-delta modulator receives an n-bit digital input signal. Before being received by the modulator, the digital input signal is typically filtered through a digital lowpass filter, typically called an interpolation filter, that also oversamples the signal from two times to 128 times the input sample rate. The filtered digital input signal is summed with the inverse feedback of the modulator's output signal to provide an error signal. The resulting error signal is then processed through an integrator and a comparator to provide an output signal. The output of sigma-delta modulator is an m-bit signal, where m is less than n. Typically, m=1, as this ensures linear conversion and is simpler to implement than an m.noteq.1 device. The one-bit digital output is typically input to an analog switched capacitor filter stage to produce an analog output. In cases where m is greater than one, a more complex analog circuit is required to convert multiple digital levels into the analog output.
Speakers and microphones typically have non-ideal frequency responses. An example of a non-ideal frequency response 100 is illustrated as having a non-flat frequency response throughout the range of relevant frequencies. A speaker having a frequency response 100, for example, would produce a distorted output, where the sound at lower frequencies is louder than the sound at higher frequencies relative to the input signal received by the speaker. The decreasing portion at the higher frequency end of the frequency response 100 is called "roll-off", and the point at which the roll-off reaches 50% of the maximum amplitude in the frequency response is called the "cutoff frequency" (at 116). An ideal frequency response is flat across all relevant frequencies and contributes no distortion.
The non-ideal frequency response of a transducer results in part from physical limitations in its design and construction. In low-cost speakers and microphones used for computer telephony and other purposes, this effect is most apparent. Nevertheless, the non-ideal frequency response of a transducer may be compensated by analog and digital filters to improve the overall frequency response of a transducer system. For example, in FIG. 1, an analog filter may be placed external to the codec 106 between the codec 106 and the speaker 102. If the frequency response of the analog filter peaks in the region corresponding to the roll-off of the transducer frequency response, the filter is said to "compensate" for the transducer's non-ideal frequency response in that region. Accordingly, the resulting effective frequency response of the entire system would more closely approximate an ideal frequency response.
Many common types of converters are capable of converting between analog and digital signals, such as flash and successive approximation converters. Such converters, however, do not typically include integral filter stages and would, therefore, require the addition of one or more compensating filters to the design to compensate for the non-ideal frequency response of a transducer. Consequently, there is a need for systems and methods that improve an overall frequency response of a transducer system without requiring additional filters. The digital lowpass filtering stages of the sigma-delta modulator discussed above have traditionally been tailored to minimize noise with very specific characteristics (e.g., at a frequency outside the pass band). However, the additional use of the same digital filtering stage to compensate for the non-ideal frequency response of the transducer, particularly the overall frequency response within the pass band, has until the present invention an unpracticed method for improving other aspects of the transducer system's performance.